stratosphere–troposphere coupling and links with...

Stratosphere–Troposphere Coupling and Links with Eurasian Land Surface Variability

JUDAH COHEN

Atmospheric and Environmental Research, Inc., Lexington, Massachusetts

MATHEW BARLOW

University of Massachusetts—Lowell, Lowell, Massachusetts

PAUL J. KUSHNER

Department of Physics, University of Toronto, Toronto, Ontario, Canada

KAZUYUKI SAITO

Frontier Research Center for Global Change, Japan Agency for Marine-Earth Science and Technology, Yokohama,Kanagawa, Japan

(Manuscript received 20 October 2006, in final form 30 January 2007)

ABSTRACT

A diagnostic of Northern Hemisphere winter extratropical stratosphere–troposphere interactions is pre-sented to facilitate the study of stratosphere–troposphere coupling and to examine what might influencethese interactions. The diagnostic is a multivariate EOF combining lower-stratospheric planetary waveactivity flux in December with sea level pressure in January. This EOF analysis captures a strong linkagebetween the vertical component of lower-stratospheric wave activity over Eurasia and the subsequentdevelopment of hemisphere-wide surface circulation anomalies, which are strongly related to the ArcticOscillation. Wintertime stratosphere–troposphere events picked out by this diagnostic often have a pre-cursor in autumn: years with large October snow extent over Eurasia feature strong wintertime upward-propagating planetary wave pulses, a weaker wintertime polar vortex, and high geopotential heights in thewintertime polar troposphere. This provides further evidence for predictability of wintertime circulationbased on autumnal snow extent over Eurasia. These results also raise the question of how the atmospherewill respond to a modified snow cover in a changing climate.

1. Introduction

In this report we examine the tropospheric precur-sors and horizontal structure of the multiple-week ex-tratropical stratosphere–troposphere interaction eventshighlighted by Baldwin and Dunkerton (1999, 2001)and numerous follow-on studies (e.g., Cohen et al.2002; Baldwin et al. 2003; Polvani and Waugh 2004;Limpasuvan et al. 2004, 2005; Reichler et al. 2005).These events consist of an annular-mode (Thompsonand Wallace 2000) signature that starts in the strato-sphere and progresses downward into the troposphere,where it persists for several weeks. The stratosphere–

troposphere annular-mode events are always precededby an anomalous lower-stratospheric planetary waveactivity flux, but robust lower-tropospheric wave activ-ity signatures of these events are more difficult to find(Polvani and Waugh 2004). This is because the tropo-sphere is inherently more variable than the strato-sphere and because the stratosphere is capable of gen-erating downward-propagating annular-mode signals—i.e., stratospheric vacillations (e.g., Holton and Mass1976; Plumb and Semeniuk 2003)—independently ofthe state of the troposphere.

The multiple-week time scale of extratropical strato-sphere–troposphere interaction events suggests thatstratospheric conditions could be used for long-rangetropospheric forecasts (Baldwin et al. 2003; Charlton etal. 2003; Siegmund 2005). If robust tropospheric pre-cursors to a significant portion of such events could be

Corresponding author address: Judah Cohen, AER, Inc., 131Hartwell Ave., Lexington, MA 02421.E-mail: [email protected]

1 NOVEMBER 2007 C O H E N E T A L . 5335

DOI: 10.1175/2007JCLI1725.1

© 2007 American Meteorological Society

JCLI4299

found, they might extend the practical forecastingrange even further back in time. With this goal in mind,we here show, following the work of Cohen et al.(2002), that these extratropical stratosphere–tropo-sphere interaction patterns can be partially predictedby conditions over the Eurasian land surface. In par-ticular, we present a concise diagnostic of the strato-sphere–troposphere interaction patterns in the North-ern Hemisphere winter (sections 2 and 3) that focuseson the spatial patterns and demonstrates that the locusof lower-stratospheric wave activity occurs over Eur-asia; we then show that these patterns are robustlylinked to variations in snow cover over Eurasia inNorthern Hemisphere autumn (section 4). It is clearthat this link does not hold for all stratosphere–tropo-sphere interaction events but does account for a signifi-cant portion of them. Besides long-range forecasts, ourfindings may have implications for understanding somerelationships associated with the global warming trend;these are discussed in the conclusions.

2. Method

Our diagnostic of the stratosphere–troposphere in-teraction patterns described in the introduction com-bines information from two physically and temporallydistinct fields. First, the December-mean vertical com-ponent of the Plumb (1985) wave activity flux (WAF)calculated from the NCEP–NCAR 40-Year Reanalysis(Kalnay et al. 1996) at 100 hPa for the Northern Hemi-sphere north of 20°N, for each year from 1948 to 2004,provides information about upward-propagating plan-etary waves. The vertical component of the WAF isgiven by

F � �p sin�2� �S�1��T� � �2�a sin�2� ��1��T��,

where is the angular velocity of rotation, p is pres-sure, is latitude, � is meridional velocity, T is tem-perature, a is the mean radius of the earth, � is geopo-tential, is longitude, and primes indicate the depar-ture from the local climatological mean. The staticstability S � ��(p/H)�T/�p � �T/H�, where anglebrackets indicate the 20°–90°N polar-cap and climato-logical average, � � R/cp and H is the scale height. Thefield F is useful for localizing in longitude and latitudethe source of vertically propagating stationary plan-etary Rossby waves. The zonal mean of F is propor-tional to the meridional eddy heat flux and the verticalcomponent of the quasigeostrophic Eliassen–Palm flux.

The second field, the January-mean NCEP–NCARReanalysis sea level pressure (SLP) for the same do-main for each year from 1949 to 2005, provides infor-mation about the lower-tropospheric circulation in the

month following the December WAF anomaly. Bothfields were each detrended, weighted by the squareroot of the cosine of the latitude, and normalized by thetheir respective mean standard deviation. Then, usingboth fields, a multivariate empirical orthogonal func-tion (EOF) analysis was conducted to identify the pri-mary patterns of common variability.

This approach is objective but heuristic: we look formodes that combine the December WAF and JanuarySLP to create a diagnostic of those strong planetarywave events that culminate with a strong SLP signal inthe troposphere. But what motivates our choice ofmonths, that is, December for the WAF and Januaryfor the SLP? The characteristic time scale of strato-sphere–troposphere interaction life cycles is severalweeks, so we expect an offset between an initial wave-driving event and the subsequent circulation response.For example, Polvani and Waugh (2004) show thatstratospheric geopotential height anomalies are corre-lated with the time-integrated lower-stratosphericWAF over the previous several weeks. To keep theanalysis reproducible and applicable to the problem ofseasonal forecasting (for which monthly anomaly fore-casts are standard products), we choose to look at WAFin one calendar month and the circulation in the sub-sequent calendar month. January is a natural targetmonth for looking at circulation anomalies for severalreasons: it is the peak month for extratropical circula-tion variability, the month of strongest annular-modecoupling between the troposphere and stratosphere(Baldwin and Dunkerton 1999; Thompson and Wallace2000), and a month of particular interest for seasonalforecasting. In section 3c, we discuss the sensitivity ofour results to the choice of analysis time period.

We calculate the leading multivariate WAF/SLPEOF and its principal component time series, shown inFig. 1. This EOF explains 15% of the combined vari-ance in the two fields (maximum local WAF correlationis 0.8, maximum local SLP correlation is 0.74) and iswell separated from the next EOF, which explains 7%of this variance. In the remainder of the article we referto the principal component of the leading multivariateWAF/SLP EOF as the stratosphere–troposphere cou-pling index (STCI), or s(t), where t is a time index run-ning from the 1948/49 to the 2004/05 December–January season. We discuss the characteristics of thismode in the next section.

A third physically distinct field we employ is a mea-sure of October-mean snow cover extent over Eurasia,which is the surface forcing that we focus on in thisstudy. This October-mean snow cover area index,which we call a(t), merges a satellite-based dataset(Robinson et al. 1993) from 1967 to 2004 and Brown’s

5336 J O U R N A L O F C L I M A T E VOLUME 20

(2000) historical reconstruction, which is based on insitu observations, from 1948 to 1966. The standardizedindices of s(t) and a(t) are shown in Table 1.

3. Results

a. A concise diagnostic of stratosphere–troposphereinteractions

In this section, we show that the multivariate EOFdescribed in section 2 efficiently captures extratropicalstratosphere–troposphere interaction events. First, weplot in Fig. 1 the WAF and SLP patterns that are as-sociated with the STCI. Figure 1a shows the correlationof December 100-hPa WAF with s(t), and Fig. 1b showsthe correlation of January SLP with s(t); covariance

produces a similar pattern for both variables (notshown). The figure represents the positive phase of themultivariate EOF, chosen by convention. We see thatthe December WAF pattern in Fig. 1a has a strongpositive center of action, corresponding to upwardwave activity propagation, over a sector extending fromeastern Europe to the east coast of Asia; this is also theregion of greatest variance, climatologically. Further-more, the center of greatest variability is over Siberia,highlighting the Eurasian sector as being particularlyimportant for stratosphere–troposphere coupling.

The January SLP (Fig. 1b) signature strongly re-sembles the negative phase of the Arctic Oscillation(AO) or Northern Annular Mode (NAM), with centersof action over the Pacific, Atlantic, and Arctic (Thomp-son and Wallace 1998, 2000). The NAM itself corre-

FIG. 1. (a) Correlation of the stratosphere–troposphere coupling index (STCI) with the vertical component of theDecember 100-hPa WAF. (b) Same as in (a), but for the correlation with January SLP. Contour interval for (a)and (b) is 0.1 starting with �0.3. (c) The STCI (solid line) and the principal component time series for the first EOFof December 100-hPa vertical WAF (dashed line) and for the first EOF of the January SLP (dashed–dotted line;this is the negative of the January Arctic Oscillation index).


sponds to the leading EOF of Northern Hemisphereextratropical January SLP, and we find that the corre-lation between the NAM index and the STCI is �0.77(Fig. 1c). This implies, for example, that an anoma-lously negative NAM phase in January is stronglylinked with an anomalously positive (upward) plan-etary WAF over Eurasia in December.

It is not surprising that the multivariate EOF is co-herent with variations in December WAF and JanuarySLP, since these fields are input to the EOF calculation.However, the fact that the multivariate EOF corre-sponds to high correlations with variability in bothfields (maximum of 0.79 in WAF and 0.77 in SLP)strongly suggests that the two fields are coupled. Inaddition, we also find that the EOF captures the time-evolving signature of stratosphere–troposphere interac-tion events. In Fig. 2a, for example, we show the cor-relation between the STCI and the zonal mean WAFarea averaged between 40° and 80°N as a function ofheight and calendar day, starting from 1 October. Theindex is coherent with the lower-stratospheric WAF inDecember, which is expected by construction. But be-yond this, the figure shows that, for example, the posi-tive phase of STCI is coherent with an upward WAFsignature that extends from the lower troposphere tothe stratosphere. In addition, there are moderate (cor-relations �0.4) but statistically significant precursors inthe lower troposphere as early as October.

Figure 2b plots the correlation between the STCI andthe polar cap (area averaged between 60° and 90°N)daily mean geopotential at each pressure level. Asmight be expected, the anomalous stratospheric wave-driving seen in Fig. 2a leads to a sign-consistent positivepolar cap geopotential anomaly—for example, a posi-tive anomaly in December WAF leads to anomalouslyhigh heights and a warmer polar stratosphere (notshown). This geopotential anomaly reaches into thetroposphere in January, leading to the polar anticy-clonic circulation anomaly seen in Fig. 1b. A more un-expected feature of this plot is the appearance of apositive height anomaly in the lower and midtropo-sphere in October and again in November; we return tothese features below.

Looking at Figs. 1 and 2 together, we see that themultivariate EOF WAF/SLP mode exhibits a multiple-week life cycle characterized by a wave activity pulseoriginating over the Eurasian sector and a subsequentzonal-mean geopotential anomaly, whose sign is con-sistent with the WAF anomaly that propagates downinto the lower troposphere. The multivariate EOF andits associated index s(t) thus efficiently pick out thoseWAF events that lead to a subsequent troposphericsignal in midwinter.

TABLE 1. Detrended standardized STCI, which is the principalcomponent of the multivariate EOF, and the Siberian Octobersnow index as described in the text for the years 1948/49–2004/05.

Year STCI Oct snow

1948 �1.59 �1.111949 �0.39 �0.591950 0.03 �0.941951 �1.09 �0.941952 �0.23 1.201953 0.06 0.321954 0.57 �1.161955 �0.27 �1.501956 �1.77 �0.441957 �0.05 �0.411958 0.27 �0.271959 1.22 1.761960 1.21 2.041961 �1.41 0.101962 0.63 0.881963 �1.12 �0.711964 0.59 0.371965 2.22 0.841966 0.21 0.561967 �0.20 �0.191968 0.98 1.131969 2.41 1.331970 0.36 1.521971 0.04 1.171972 �0.83 1.121973 �0.04 0.241974 0.03 �0.071975 �0.68 �0.691976 2.25 3.451977 �0.40 0.841978 0.87 0.231979 �0.28 �0.691980 �1.50 �1.101981 1.30 �0.531982 0.48 0.531983 �0.17 �0.261984 1.99 �0.131985 �0.40 �0.121986 �0.96 �0.481987 �0.07 �1.291988 �1.78 �1.901989 �1.34 �0.281990 �0.62 �1.001991 �0.81 �1.161992 �1.45 �0.581993 1.34 0.371994 �0.50 �1.101995 �0.72 0.031996 0.22 0.201997 1.22 �0.221998 0.23 1.001999 �0.58 0.052000 0.15 0.342001 �0.10 0.672002 �0.02 1.852003 1.09 0.402004 �0.59 0.66


Why is the lower-stratospheric WAF EOF structurein Fig. 1a focused over the Eurasian sector? Since theclimatological maximum in the vertical component ofthe Plumb flux is located there (Plumb 1985, his Fig. 4b)it might be expected that this region would also repre-sent the area of maximum variability of this quantity,which the EOF would then identify. Plumb discusseshow wave activity fluxes in this sector are forced byorography and diabatic heating, and presumably varia-tions in these forcings contribute to the EOF pattern inFig. 1a. In the next section, we discuss how one sourceof diabatic heating variability, related to Eurasian snowcover, might account for some portion of the WAFpattern in Fig. 1a and for the subsequent transientstratosphere–troposphere interactions.

b. October Eurasian snow

We now pursue the suggestion of Cohen and othersthat strong anomalies in October Eurasian snow covercan force a large-scale circulation response that in-volves a stratosphere–troposphere interaction event.The proposed mechanism is that high-albedo snowanomalies occurring sufficiently early in the fall canlead to a strong shortwave radiation anomaly at thesurface and to a local diabatic-heating anomaly. Thequestion then is whether this can excite a planetaryRossby wave response that sets off a stratosphere–troposphere interaction event. If the forcing is on aregional scale, the scale of the Rossby wave responsewould become independent of the scale of the forcingand reflect instead the detailed zonal mean-flow param-eters as determined by the dynamics of planetaryRossby wave propagation [see, e.g., the stationary waveresponse to a narrow topographic ridge, discussed inHeld et al. (2002)].

We find a good deal of statistical evidence to supportthe hypothesis that October Eurasian snow extent canpredict stratosphere–troposphere interactions. First ofall, the STCI, s(t), and the October Eurasian snow in-dex a(t) are correlated at 0.56, which is statistically sig-nificant at the confidence limit greater than 99% basedon the Student’s t test. The scatterplot and associatedlinear fit of these two quantities are shown in Fig. 3.This associates a fraction of positive October snowevents with an upward-propagating planetary wavepulse in December and a negative NAM response inthe stratosphere and troposphere into January.

For further evidence of the predictive power of Eur-

FIG. 2. Correlation of the principal component of multivariateEOF shown in Fig. 1 with (a) daily vertical wave activity fluxaveraged for all grid boxes between 40° and 80°N and all verticallevels between 850 and 10 hPa, and (b) daily geopotential heightaveraged for all grid boxes north of 60°N and all vertical levelsbetween 1000 and 10 hPa. Those values exceeding 90%, 95%, and99% confidence intervals based on Student’s t test are denoted bylight, dark, and darkest shading, respectively.

FIG. 3. Scatterplot of the principal component or time seriesassociated with multivariate EOF shown in Fig. 1 and EurasianOctober snow cover for the years 1948–2004. Correlation betweenthe two time series is 0.56.


Fig 2 live 4/C

asian snow cover, we look to the time-evolving signa-ture of the WAF and of the circulation that is coherentwith the snow index a(t). Figure 4a plots the day-by-daycorrelation of the October snow index a(t) with theareal-mean WAF, and Fig. 4b plots the correlation ofa(t) with the polar cap mean geopotential height. Thatis, Fig. 4 is the same as Fig. 2, except that October snowcover area is substituted for the STCI. We see here thatthe October snow index is coherent with planetarywave pulses that originate in the troposphere in No-vember and with a strong pulse that propagates fromthe lower troposphere into the stratosphere in Decem-ber. Furthermore, we see a similar pattern in the geo-potential correlations in Fig. 4b to what was seen in Fig.2b; the resemblance and the fact that a vertically co-

herent correlation between October snow and the ex-tratropical circulation emerges in December and sur-vives into January is remarkable.

In Fig. 4 we see that the correlations between Octo-ber snow cover and the atmospheric circulation are notgreatest during October but almost 3 months later, inJanuary. This suggests that the high correlations be-tween snow cover and winter climate are not simplydue to atmospheric conditions in October and thatsnow cover does not act simply as a reservoir of atmo-spheric memory. Figures 2 and 4 suggest that a series ofrelatively weak WAF pulses act to precondition thestratosphere prior to the main pulse. We cannot explainthe timing of the individual WAF pulses through thefall–winter season but expect that these pulses are tiedto the positive NH tropospheric polar cap heightanomalies in fall seen in Figs. 2 and 4. This connectionis suggested by previous work of Cohen and others (seeCohen et al. 2001, 2002) that links positive lower-tropospheric geopotential height anomalies over Eur-asia in fall to negative NAM events in the tropospherein winter.

c. Sensitivity of results to analysis time period

The restricted December/January analysis period,and the fact that we have retained only the leadingmode of the multivariate EOF, means that our STCIcannot capture all the stratosphere–troposphere cou-pling events. In particular, the North Atlantic sectorrepresents another source region for vertical WAFfrom the troposphere into the stratosphere (Plumb1985), and not all STC events will result in NAM cir-culation signatures, and many STC events occur outsidethe December/January time period (Baldwin andDunkerton 1999; Thompson and Wallace 2000).

In this subsection, we focus on how sensitive our re-sults might be to a change of analysis time period. Weproceed as follows:

1) We compute the first EOF of the vertical compo-nent of the WAF at 100 hPa for the months using atime series of monthly averaged WAF as a functionof latitude and longitude from 40° to 80°N, includingthe autumn-to-winter months (September–January)only. The leading mode, shown in Fig. 5a, closelyresembles the pattern shown in Fig. 1a, with themain center of variability over northern Eurasia.

2) We then repeat step 1 but for the SLP as a functionof latitude and longitude from 20° to 90°N, in a timeperiod lagged by 1 month from the WAF, that is,for the months October–February. The leadingmode, shown in Fig. 5b, appears as a classical AOpattern, as in Fig. 1b.

FIG. 4. Correlation of the areal extent of October Eurasiansnow cover with (a) daily vertical wave activity flux averaged forall grid boxes between 40° and 80°N and all vertical levels be-tween 850 and 10 hPa, and (b) daily geopotential height averagedfor all grid boxes north of 60°N and all vertical levels between1000 and 10 hPa. Those values exceeding 90%, 95%, and 99%confidence intervals based on Student’s t test are denoted by light,dark, and darkest shading, respectively.


Fig 4 live 4/C

3) We form daily time series indices for the patterns inFigs. 5a and 5b by calculating the projection of thesepatterns onto the daily WAF and SLP, using 1 Sep-tember–31 January for the WAF and 1 October–28February for the SLP. We plot the lagged temporalcorrelation between these two time series in Fig. 5c;only statistically significant points are shaded, andlines have been drawn to delineate the months.

In Fig. 5c, the greatest percentage of statistically sig-nificant correlations occurs between December WAFand January SLP. This further supports our choice of

time period, and the similarity between Figs. 5a,b andFigs. 1a,b emphasizes the overall robustness of thecoupled mode of variability we have identified. Inter-estingly, snow cover could be thought of as an indepen-dent variable corroborating these results. As seen fromFig. 4, correlations between October snow cover andWAF are highest in December, and those between Oc-tober snow cover and geopotential heights are highestin January.

The region with the second greatest percentage ofstatistically significant correlations in Fig. 5c is that ofNovember WAF and January SLP. This is consistent

FIG. 5. (a) The first EOF of 100-hPa vertical wave activity flux for the 5-month period September–January shownas a correlation between the principal component and the WAF. (b) The first EOF of SLP for the 5-month periodOctober–February shown as a correlation between the principal component and the SLP. Contour interval for (a)and (b) is 0.1 starting with �0.1. (c) Daily lead-lag correlations for the two EOFs shown in (a) and (b) projectedon to the daily data.


Fig 5 live 4/C

with Fig. 2a and with the results of Polvani and Waugh(2004), who found that the dominant mode of geopo-tential height variability in the stratosphere was highlycorrelated with the vertical WAF at 100 hPa from theprior 6 weeks. So if a NAM event originates in thestratosphere and propagates down to the surface inJanuary, upward WAF would contribute to the eventnot only from December but from November as well.

Because of the patchiness of the significance regionsin Fig. 5c, we assess, using a Monte Carlo test, whetherthe locally statistically significant regions identified bythe t test were indeed robust. In this test, we generate1000 random time series with the same autocorrelationstatistics as the WAF and AO time series and computethe percent area of the lag correlation found to be sta-tistically significant above the 95% threshold. Thisanalysis shows that only the statistically significant re-gion of correlations between the December WAF andJanuary AO is robust. Thus, although the results pre-sented in Fig. 5c suggest a more extended period ofstratosphere–troposphere coupling throughout thefall and winter months, our limited analysis highlightsDecember and January as the preferred months forupward WAF and the resulting tropospheric AOanomaly.

4. Conclusions

To summarize our findings, we have developed aconcise diagnostic of stratosphere–troposphere interac-tions that captures those strong lower-stratosphericplanetary wave activity events that later connect to thesurface. The locus of the lower-stratospheric wave ac-tivity is over Eurasia with the center over Siberia, col-located with the climatological maximum of variance.Also, the resultant atmospheric response at the surfaceis closely related to the AO or NAM pattern of vari-ability. We further show that this type or mode of cou-pling is favored in the December to January time frame.

We have found that the STCI arising from this analy-sis is coherent with October Eurasian snow extent,which reveals a connection between October Eurasiansnow cover and stratosphere–troposphere interactionevents. Typical correlations, shown in Fig. 4, are mod-erate (�0.5) but significant, suggesting that a significantportion of the anomalous snow events in October aretied to stratosphere–troposphere circulation anomaliesmore than 8 weeks later. The observational results pre-sented here linking variability in snow cover withstratosphere–troposphere coupling are consistent withthe modeling results of Gong et al. (2002). This studyshowed that interannual variations in snow cover arerequired to capture the observed significant correla-

tions between the tropospheric NAM and the strato-spheric NAM.

We outline in Fig. 6 (based on Saito 2003; see alsoReichler et al. 2005, their Fig. 1) a conceptual model ofthe dynamical pathway demonstrated by the statisti-cally significant relationship among all the variables dis-cussed above. October is the month snow cover makesits greatest advance, mostly across Siberia. October isalso the month in which the Siberian high, one of thethree dominant centers of action across the NH, forms.In years when snow cover is above normal this leads toa strengthened Siberian high and colder surface tem-peratures across Northern Eurasia in the fall. We sug-gest that the intensification of the Siberian high, alongwith the thermal impacts of enhanced snow cover andtopographic forcing, corresponds to a positive wave ac-tivity flux anomaly in the late fall and early winter,leading to stratospheric warming and to the Januarytropospheric negative AO response we have described.

There are important implications of this work forseasonal forecasting. We have described here a statis-

FIG. 6. Conceptual model for how fall snow cover modifieswinter circulation in both the stratosphere and the troposphere;case for extensive snow cover illustrated: 1) Snow cover increasesrapidly in the fall across Siberia, when snow cover is above nor-mal. 2) Diabatic cooling helps strengthen the Siberian high andleads to below normal temperatures. 3) Snow-forced diabaticcooling in proximity to the high topography of Asia increasesupward flux of wave activity from the troposphere, which is ab-sorbed in the stratosphere. 4) Strong convergence of WAF leadsto higher geopotential heights, a weakened polar vortex, andwarmer temperatures in the stratosphere. 5) Zonal mean geopo-tential height and wind anomalies propagate down from thestratosphere into the troposphere all the way to the surface. 6)Dynamic pathway culminates with strong negative phase of theArctic Oscillation at the surface.


tically significant relationship between time-leading ex-tratropical boundary condition anomalies and time-lagging extratropical circulation anomalies. Further-more, we have provided a diagnostic that can be used asan aid in operational monthly to seasonal forecasts.This complements ENSO-based predictors and sug-gests that there is hope for improving seasonal forecastsfrom extratropical indicators. Indeed, winter forecaststhat incorporate snow cover as one of its predictorshave demonstrated greater skill than forecasts that relyexclusively on ENSO (Cohen and Fletcher 2007).

We are also exploring the implications of this workfor longer-term climate trends. If fall snow extent forcesa consistent NAM response in winter, through the com-bination of diabatic and planetary wave generationmechanisms as we have described, then we expect thatlong-term trends in snow extent will be linked to long-term NAM variability and trends. Initial work alongthese lines (Cohen and Barlow 2005) demonstrates thatthere are in fact statistical connections between long-term trends in Eurasian snow cover and the tropo-spheric circulation. We are currently building on theCohen–Barlow results with a focus on the stratosphere–troposphere interactions we have discussed here.

Acknowledgments. Cohen was supported by NSFGrant 0443512, Barlow by NSF Grant 0603555, andKushner by the Canadian Foundation for Climate andthe Atmospheric Sciences. We thank the efforts of tworeviewers whose comments improved the quality of theoriginal manuscript.

REFERENCES

Baldwin, M. P., and T. J. Dunkerton, 1999: Propagation of theArctic Oscillation from the stratosphere to the troposphere.J. Geophys. Res., 104, 30 937–30 946.

——, and ——, 2001: Stratospheric harbingers of anomalousweather regimes. Science, 294, 581–584.

——, D. B. Stephenson, D. W. J. Thompson, T. J. Dunkerton,A. J. Charlton, and A. O’Neill, 2003: Stratospheric memoryand extended-range weather forecasts. Science, 301, 636–640.

Brown, R. D., 2000: Northern Hemisphere snow cover variabilityand change, 1915–97. J. Climate, 13, 2339–2355.

Charlton, A. J., A. O’Neill, D. B. Stephenson, W. A. Lahoz, andM. P. Baldwin, 2003: Can knowledge of the state of thestratosphere be used to improve statistical forecasts of thetroposphere? Quart. J. Roy. Meteor. Soc., 129, 3205–3224.

Cohen, J., and M. Barlow, 2005: The NAO, the AO, and globalwarming: How closely related? J. Climate, 18, 4498–4513.

——, and C. Fletcher, 2007: Improved skill of Northern Hemi-sphere winter surface temperature predictions based onland–atmosphere fall anomalies. J. Climate, 20, 4118–4132.

——, K. Saito, and D. Entekhabi, 2001: The role of the Siberianhigh in Northern Hemisphere climate variability. Geophys.Res. Lett., 28, 299–302.

——, D. Salstein, and K. Saito, 2002: A dynamical framework tounderstand and predict the major Northern Hemispheremode. Geophys. Res. Lett., 29, 1412, doi:10.1029/2001GL014117.

Gong, G., D. Entekhabi, and J. Cohen, 2002: A large-ensemblemodel study of the wintertime AO–NAO and the role ofinterannual snow perturbations. J. Climate, 15, 3488–3499.

Held, I. M., M. Ting, and H. Wang, 2002: Northern winter sta-tionary waves: Theory and modeling. J. Climate, 15, 2125–2144.

Holton, J. R., and C. Mass, 1976: Stratospheric vacillation cycles.J. Atmos. Sci., 33, 2218–2225.

Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.

Limpasuvan, V., D. W. J. Thompson, and D. L. Hartmann, 2004:The life cycle of Northern Hemisphere sudden stratosphericwarmings. J. Climate, 17, 2584–2596.

——, D. L. Hartmann, D. W. J. Thompson, K. Jeev, and Y. L.Yung, 2005: Stratosphere-troposphere evolution duringpolar vortex intensification. J. Geophys. Res., 110, D24101,doi:10.1029/2005JD006302.

Plumb, R. A., 1985: On the three-dimensional propagation of sta-tionary waves. J. Atmos. Sci., 42, 217–229.

——, and K. Semeniuk, 2003: Downward migration of extratro-pical zonal wind anomalies. J. Geophys. Res., 108, 4223,doi:10.1029/2002JD002773.

Polvani, L. M., and D. W. Waugh, 2004: Upward wave activityflux as a precursor to extreme stratospheric events and sub-sequent anomalous surface weather regimes. J. Climate, 17,3548–3554.

Reichler, T., P. J. Kushner, and L. M. Polvani, 2005: The coupledstratosphere–troposphere response to impulsive forcing fromthe troposphere. J. Atmos. Sci., 62, 3337–3352.

Robinson, D. A., K. F. Dewey, and R. R. Heim Jr., 1993: Globalsnow cover monitoring: An update. Bull. Amer. Meteor. Soc.,74, 1689–1696.

Saito, K., 2003: Seasonal to sub-decadal climatic covariability ofsnow cover and atmosphere in the Northern Hemisphere.Ph.D. thesis, Nagoya University, 161 pp.

Siegmund, P., 2005: Stratospheric polar cap mean height and tem-perature as extended-range weather predictors. Mon. Wea.Rev., 133, 2436–2448.

Thompson, D. W. J., and J. M. Wallace, 1998: The Arctic Oscil-lation signature in the wintertime geopotential height andtemperature fields. Geophys. Res. Lett., 25, 1297–1300.

——, and ——, 2000: Annular modes in the extratropical circula-tion. Part I: Month-to-month variability. J. Climate, 13, 1000–1016.


stratosphere–troposphere coupling and links with...

Documents